Chem. Res. Toxicol. 2008, 21, 421–431
421
Binding and Hydrolysis of Soman by Human Serum Albumin Bin Li,† Florian Nachon,‡,§ Marie-Thérèse Froment,‡,§ Laurent Verdier,# Jean-Claude Debouzy,‡,4 Bernardo Brasme,‡,⊥ Emilie Gillon,‡,§ Lawrence M. Schopfer,† Oksana Lockridge,† and Patrick Masson*,‡,§ UniVersity of Nebraska Medical Center, Eppley Institute, Omaha, Nebraska 68198-6805, Centre de Recherches du SerVice de Santé des Armées, Département de Toxicologie, Unité d’Enzymologie, Unité de Biophysique, SerVice de Biospectroscopie, BP 87, 38702 La Tronche cédex, France, 3, and Département Analyse Chimique, Centre d’Etudes du Bouchet, BP 3, 91710 Vert-le-Petit, France ReceiVed September 15, 2007
Human plasma and fatty acid free human albumin were incubated with soman at pH 8.0 and 25 °C. Four methods were used to monitor the reaction of albumin with soman: progressive inhibition of the aryl acylamidase activity of albumin, the release of fluoride ion from soman, 31P NMR, and mass spectrometry. Inhibition (phosphonylation) was slow with a bimolecular rate constant of 15 ( 3 M-1 min-1. MALDI-TOF and tandem mass spectrometry of the soman-albumin adduct showed that albumin was phosphonylated on tyrosine 411. No secondary dealkylation of the adduct (aging) occurred. Covalent docking simulations and 31P NMR experiments showed that albumin has no enantiomeric preference for the four stereoisomers of soman. Spontaneous reactivation at pH 8.0 and 25 °C, measured as regaining of aryl acylamidase activity and decrease of covalent adduct (pinacolyl methylphosphonylated albumin) by NMR, occurred at a rate of 0.0044 h-1, indicating that the adduct is quite stable (t1/2 ) 6.5 days). At pH 7.4 and 22 °C, the covalent soman-albumin adduct, measured by MALDI-TOF mass spectrometry, was more stable (t1/2 ) 20 days). Though the concentration of albumin in plasma is very high (about 0.6 mM), its reactivity with soman (phosphonylation and phosphotriesterase activity) is too slow to play a major role in detoxification of the highly toxic organophosphorus compound soman. Increasing the bimolecular rate constant of albumin for organophosphates is a protein engineering challenge that could lead to a new class of bioscavengers to be used against poisoning by nerve agents. Soman-albumin adducts detected by mass spectrometry could be useful for the diagnosis of soman exposure. 1. Introduction Albumin is an abundant protein that represents 50–60% of the total protein in human plasma and body fluids. Its concentration in plasma is about 0.6 mM. Albumin displays both an esterase activity (1) and an aryl acylamidase activity (2, 3). The topology of the esterase/amidase active site of albumin has been probed by site-directed mutagenesis (4) and X-ray structure determination of several drug-albumin complexes (5). Tyr411 was determined to be the catalytic nucleophile in these reactions. Albumin is also known to bind organophosphates (OPs1) and carbamates (6–8) and to react with them (9–11). OPs and certain carbamates are actually hydrolyzed by albumin through transient phosphylation/alkylation of its active site (12–17). The residue that reacts with OPs was proven to be a tyrosine (9-11, 14, 18). DF32P-labeling of human albumin followed by peptide sequencing showed that the labeled tyrosine is in the tetrapeptide * To whom correspondence should be addressed. Patrick Masson, CRSSA, Département de Toxicologie, Unité d’Enzymologie, BP 87, 38702 La Tronche cédex, France. Tel: +33 (0)4 76 63 69 59. Fax: +33 (0)4 76 63 69 62. E-mail:
[email protected]. † Eppley Institute. ‡ Centre de Recherches du Service de Santé des Armées. § Département de Toxicologie, Unité d’Enzymologie. # Centre d’Etudes du Bouchet. 4 Unité de Biophysique. ⊥ Service de Biospectroscopie. 1 Abbreviations: AAA, aryl acylamidase; AChE, acetylcholinesterase; BuChE, butyrylcholinesterase; DFP, diisopropylfluorophosphate; FAF-HSA, fatty acid-free human serum albumin; HSA, human serum albumin; MP, methylphosphonate; PMP, pinacolyl methylphosphonate; o-NTFNAC, N-(onitrophenyl)trifluoroacetamide; OP, organophosphorus ester.
sequence Arg-Tyr-Thr-Lys (9). Arg and Tyr were subsequently identified as Arg 410 and Tyr 411, the key residues of the esteratic site of HSA (4). Recently, DFP-inhibition of the aryl acylamidase (AAA) activity of FAF-HSA confirmed that Tyr 411 is also the nucleophilic pole of the albumin AAA activity (2). Finally, MALDI-TOF mass spectrometry provided direct evidence that the OPs chlorpyrifos-oxon, dichlorvos, DFP, and sarin bind covalently to human albumin at Tyr 411 (19). In the present work, we investigated the reaction of human albumin with soman. MALDI-TOF and quadrupole tandem MS/ MS mass spectrometry of the soman-albumin adduct showed that Tyr411 was phosphonylated. Unlike soman adducts of butyrylcholinesterase, the albumin-soman adduct did not age, that is, lose its pinacolyl chain. Covalent docking simulations and 31P NMR experiments indicated that there was no enantiomeric preference of albumin for the stereoisomers of soman. Kinetic parameters of albumin phosphonylation by soman and subsequent dephosphonylation were determined. Our results could have application for the detection of soman exposure in humans. Mass spectrometry could be used to detect soman-albumin adducts. In addition, antibodies to the somanalbumin adduct could be generated for use in a rapid antibodybased assay of soman exposure. Lastly, mutants of albumin could lead to a new class of scavengers against OP poisoning.
2. Experimental Procedures 2.1. Chemicals. Fatty acid-free human albumin (FAF-HSA) and porcine pepsin were from Sigma Chemical Co. (Saint Quentin
10.1021/tx700339m CCC: $40.75 2008 American Chemical Society Published on Web 12/29/2007
422 Chem. Res. Toxicol., Vol. 21, No. 2, 2008 Fallavier, France). o-NTFNAC was a gift from Dr. Sultan Darvesh (Dalhousie University, Halifax, Canada). Racemic soman (pinacolyl methylfluorophosphate) dissolved in isopropanol was from CEB (Vert-le-Petit, France). All other chemicals were of biochemical grade. The concentration of racemic soman in the stock solution was determined by programmed temperature gas chromatography after hydrolysis into methylpinacolyl phosphonic acid and derivatization with pentafluorobenzyl bromide (20). It was 4.73 mg/mL (26 mM). 31 P NMR spectra of 2 mM soman in deuterated dimethylsulfoxide were recorded at 27 °C on a Bruker spectrometer (AM400, 9.4 T). The spectra showed that the four stereoisomers are present in nearly equimolar amounts. Soman has two chiral centers. The absolute configuration of the four diastereoisomers is known. PS, PR, CS, and CR correspond to the old notation P- P+, C+, and C-, respectively. Thus, isomers PSCS ) P-C+, PSCR ) P-C-, PRCR ) P+C-, and PRCS ) P+C+. Stereoisomers PSCS and PSCR are the most active toward the biological target acetylcholinesterase (21). 2.2. Aryl Acylamidase Enzymatic Assay of Albumin. The AAA activity of albumin was assayed with o-NTFNAC (2 mM) as the substrate in 60 mM Tris/HCl buffer at pH 8.0 at 25 °C. A 60 mM stock solution of o-NTFNAC was prepared in 50% water/ acetonitrile (v/v). Because isopropanol is the solvent for soman, isopropanol was also added to the buffer in the control assay for the AAA activity of albumin in the absence of soman. The final concentration of acetonitrile in the assay was 3.3% and that of isopropanol was 2%. The release of the phenolic product (onitroaniline) was monitored for 5 min at 430 nm (ε ) 3954 M-1 cm-1) according to Darvesh et al. (22). Because of the low enzymatic activity of albumin with o-NTFNAC (3), high concentrations of FAF-HSA were used in assays (0.075 mM final). Measured rates were corrected for spontaneous hydrolysis of o-NTFNAC. 2.3. Mass Spectrometry (MALDI-TOF and Quadrupole MS/MS) of the Soman-Albumin Adduct from Human Plasma. One hundred microliters of human plasma was mixed with 2.3 µL of stock soman (26 mM in isopropanol) to give a reaction mixture containing 600 µM soman (in 2.3% isopropanol). The mixture was incubated at room temperature for 3–7 days before processing. No buffer was added to the plasma during this step. The pH of a 10 µL aliquot was reduced to pH 2.3 by adding 10 µL of 1% trifluoroacetic acid. Proteins were digested with 0.5 µg of pepsin for 2 h at 37 °C. At pH 2.3, selective proteolysis at the C-terminal side of the leucine and phenylalanine residues is expected. There was no need to denature the proteins or to reduce and alkylate the disulfide bonds because the peptides of interest were released without these added steps. Peptides were separated on a C18 reverse phase column on a Waters 625 LC system with a 40 min gradient starting with 85% buffer A (0.1% trifluoroacetic acid in water), 15% buffer B (acetonitrile containing 0.07% trifluoroacetic acid), and ending with 65% buffer A and 35% buffer B. One milliliter fractions were reduced in volume to 200 µL in a vacuum centrifuge, and 1 µL was analyzed by MALDI-TOF with a 2,5-dihydroxybenzoic acid matrix. Mass spectra were acquired with the Applied Biosystems Voyager DE-PRO MALDI-TOF mass spectrometer in linear positive ion mode. The spectrometer was calibrated using standard peptides (Applied Biosystems Mixture 1). A control plasma sample was identically treated, except that it was incubated with 3.1% isopropanol rather than with soman. MS/MS spectra were acquired on a Q-Trap 2000 triple quadrupole linear ion trap mass spectrometer (Applied Biosystems, MDS Sciex, Foster City, CA) with a nano electrospray ionization source. Samples were infused into the mass spectrometer at 0.35 µL/min via a fused silica emitter (360 µm o.d., 20 µm i.d., 15 µm taper, New Objective, Woburn, MA) using a Harvard syringe pump to drive a 25 µL Hamilton syringe equipped with an inline 0.25 µm filter. Samples were sprayed with 50% acetonitrile and 0.1% formic acid. Mass spectra were calibrated using fragment ions generated from collision-induced dissociation of Glu fibrinopeptide B (Sigma). Enhanced product ion scans were obtained with a collision energy of 50 ( 5 V and a pure nitrogen gas pressure of 4 × 10-5 Torr.
Li et al. The final enhanced product ion scan was the average of 105 scans. Ions were identified by manual sequencing. 2.4. 31P NMR Spectroscopy. 31P NMR spectra for the reaction between albumin (0.78 mM or 1.3 mM) and soman (1.3 mM) in 60 mM Tris/HCl buffer at pH 8.0 were recorded at 27 °C on a Bruker AM400NB spectrometer operating at 162 MHz. The spectra were acquired using successive 8 acquisition blocks of 5,000 accumulated scans (acquisition time: 3 h/spectrum) using 30 kHz spectral width, 32,000 acquisition points, and a composite pulse proton decoupling (CPD mode). The external reference for chemical shifts was 85% (w/v) H3PO4. Spectra were compared to the spectrum of soman in buffer and the spectrum of methyl pinacolyl phosphonate (MPP) in the presence and absence of albumin. MPP was made by the complete hydrolysis of soman in 5 N sodium hydroxide.
2.5. Kinetic Studies of the Reaction of Albumin with Soman. 2.5.1. Residual AAA Activity. FAF-HSA (780 µM) was incubated with 120 to 1300 µM concentrations of racemic soman in 60 mM Tris/HCl buffer at pH 8.0 at 25 °C. The time dependence of the inhibition of albumin by soman was monitored by following the residual AAA activity of albumin. Measurements of the AAA residual activity were performed on 100 µL aliquots of reaction mixture, using the sampling method (23). 2.5.2. Monitoring the Release of Fluoride from Soman. The release of fluoride, that is, the leaving group of soman, upon the reaction of soman with albumin and spontaneous hydrolysis was monitored by ionometry using a thermostatted ionometer (Radiometer IONcheck 45) equipped with an ion-selective electrode for fluoride (ISE 301F). FAF-HSA (780 µM) was incubated with 520, 780, and 1300 µM concentrations of racemic soman in 60 mM Tris/HCl buffer at pH 8.0 at 25 °C, and the concentration of released fluoride was assayed. The spontaneous hydrolysis of soman was determined under the same conditions. 2.5.3. Kinetic 31P NMR Spectroscopy. To follow the phosphonylation reaction, 1.33 mM FAF-HSA was incubated with 1.35 mM soman in 60 mM Tris/HCl buffer at pH 8.0 at 25 °C in the presence of deuterated DMSO (10% v/v final). The successive spectra were recorded every 20 min (800 scans) over 4 h. Concentrations of the different species were plotted as a function of time. To follow the reactivation of phosphonylated albumin, 1.3 mM FAF-HSA was reacted with one equivalent of racemic soman in 60 mM Tris/HCl buffer at pH 8.0. Then, the phosphonylated albumin solution was maintained at 25 °C, and spectra were recorded every 2 h over 60 h. Kinetic constants for the spontaneous hydrolysis of soman, phosphonylation of albumin, and reactivation of albumin (dephosphonylation) were determined by fitting data against the numeric solution of differential equations that describe Scheme 1, using Mathematica 5 (Wolfram Research). 2.5.4. Kinetic Mass Spectrometry. The stability of the covalent soman-albumin adduct was measured on a MALDI-TOF-TOF 4800 mass spectrometer (Applied Biosystems). Human albumin was labeled by incubating a 1 mg/mL solution (15 µM) in 10 mM Tris/ HCl at pH 8.0 with 200 µM soman for 24 h at room temperature. After this time, the concentration of active soman was negligible as determined by measuring the inhibition of human butyrylcholinesterase. The albumin solution was diluted to 0.1 mg/mL with 0.01% sodium azide in water and the pH adjusted to 7.4. The 200 µL solution was stored at 22 °C. Every 2 or 3 days, a 10 µL aliquot was acidified with 10 µL of 1% trifluoroacetic acid and digested with 2 µL of 1 mg/mL porcine pepsin (dissolved in 10 mM HCl). The digest was incubated at 37 °C for 1–4 h, and then 0.5 µL was spotted on a MALDI target plate. The dry spot was overlaid with 0.5 µL of 10 mg/mL R-cyano-4-hydroxycinnamic acid in 50% acetonitrile and 0.1% trifluoroacetic acid. MS scans in reflector mode were acquired at 3000 V by summing 500 laser shots per scan. The percent label on Tyr 411 was calculated from cluster areas. Peptide masses acquired in reflector mode are monoisotopic, which makes their mass about 1 amu lower than the average mass acquired in linear mode.
Reaction of Albumin with Soman
2.6. Molecular Modeling. 2.6.1. Noncovalent Docking. Docking calculations were carried out using two programs: (1) Autodock, version 3.0.5, with the Lamarckian genetic algorithm (LGA (24)) and (2) Gold, version 3.1 (CCDC Software, Ltd). Docking with Autodock employed the following procedure. The molecular models of the four diastereoisomers of soman were built and minimized with the MM2 force field of Chem3D (Cambridge soft.). The structure of HSA was prepared from the crystal structure of its complex with indoxyl sulfate (pdb code 2bxh) or warfarin (pdb code 2bxd) to take into account the heterogeneity of conformation of Val433. Molecules of water and ligands were removed from the model. Soman and HSA were further prepared using Autodock Tools 1.4 (25). The 3D affinity grid box was designed to include the full pocket near Tyr411. The number of grid points in the x-, y-, and z-axes was 60, 60, and 60 with grid points separated by 0.375 Å. Docking calculations were set to 100 runs. At the end of the calculation, Autodock performed cluster analysis. Docking solutions with ligand all-atom root-mean-square deviation (rmsd) within 1.0 Å of each other were clustered together and ranked by the lowest energy representative. The lowest-energy solution was accepted as the one most representative of the soman-HSA complex. Docking with Gold employed the following procedure. Protein structures were protonated and minimized by conjugate gradients using Gromacs 3.3 (26, 27) and the parameter set 53A6 (28). Ligands were built into the Sybyl 7.2 package (Tripos Inc.). The Mopac-PM3 (29) semiempirical method was chosen for geometry optimization and calculation of atomic charges. The cavity was defined as residues having atoms up to 15 Å from the Tyr411 hydroxyl group. We used long search settings for the Gold GA: 40 docking with parameters corresponding to the default 200% search efficiency settings. The final selection of conformers was based on the two scoring functions, Goldscore (30) and Chemscore (31). 2.6.2. Covalent Docking. This simulation was carried out using Gold, version 3.1 with GA settings similar to those for noncovalent docking. We used the same proteins as before, and the results were scored using both Goldscore and Chemscore functions. The fluorine atom from each of the four soman stereoisomers was replaced by a hydroxyl group, and the phosphorus stereochemistry was inverted to get the right configuration when the phosphorus atom binds to the Tyr411 Oγ. The highest score isomers were chosen as the ones giving the lowest ∆Hf when estimating the energy of the system with Mopac-PM3. To reduce the number of atoms in our system, we removed residues having all atoms further than 5 Å from any soman atoms.
3. Results 3.1. Mass Spectrometry to Identify the Covalent Binding of Soman to Tyrosine 411. Peptic digests of soman-labeled and control human plasma were separated by HPLC. Matched fractions from the two preparations were examined by MALDITOF mass spectrometry. MALDI-TOF signals for soman-albumin and control albumin peptides are shown in Figure 1. A peptide with mass 1880 amu is present in fraction 21 from the somanlabeled preparation (Figure 1, panel B). This mass is consistent with the peptide (L)VRY*TKKVPQVSTPTL(V) (1718 amu, from residues 409–423 of the mature albumin sequence) with an added mass of 162 amu from soman. A peptide with mass 1993 amu is present in fraction 23 from the soman-labeled preparation (Figure 1, panel D). This mass is consistent with the peptide (L)LVRY*TKKVPQVSTPTL(V) (1831 amu, from residues 408–423 of the mature albumin sequence, representing a missed peptic cleavage) with an added mass of 162 amu from soman. The parallel fractions from control human plasma do not have peptides of these masses (Figure 1, panels A and C). However, other prominent peaks are present at identical masses in both the control and soman-labeled spectra (compare panels A to B, and C to D), confirming that the fractions are matched with regard to HPLC elution and relative mass spectral
Chem. Res. Toxicol., Vol. 21, No. 2, 2008 423
Figure 1. MALDI-TOF mass spectra of peptic peptides from the soman-albumin adduct. Panels A and C, control human plasma digested with pepsin and fractionated by HPLC, but not treated with soman. Panels B and D, human plasma treated with soman before digestion with pepsin and fractionation by HPLC. Soman-labeled albumin peptides of 1880 and 1993 amu include 162 amu from soman.
sensitivity. Tyr411 is present in both peptides (indicated by the star), supporting the proposal that soman binds covalently to Tyr411 of human albumin in plasma. Peptides with an added mass of 162 from soman, but not with an added mass of 78 were found. This indicates that the soman adduct does not lose the pinacolyl group when soman is bound to albumin. Thus, the soman-albumin adduct does not age. An additional experiment was performed to demonstrate by a second mass spectrometry method that soman was covalently bound to Tyr 411 of human albumin. HPLC fraction 21 containing the soman-labeled albumin peptide at 1880 m/z was infused into the Q-Trap, quadrupole mass spectrometer. The enhanced mass spectrum showed a peak at 940.7 m/z, which is consistent with the doubly charged form of the 1880 peptide. This peptide was subjected to collision-induced dissociation. The resulting enhanced product ion spectrum yielded amino acid sequence information consistent with the sequence VRYTKKVPQVSTPTL, with soman covalently bound to tyrosine (see Figure 2). The prominent peak at 898.6 m/z is a doubly charged ion, 42 mass units less than the parent ion at 940.7 m/z. This is consistent with the loss of the pinacolyl group (84 mass units) from the doubly charged parent to form a doubly charged product retaining methylphosphonic acid. A comparable 42 amu loss from the parent ion was seen in the enhanced product ion spectrum of the LVRYTKKVPQVSTPTL peptide (data not shown). Preferential loss of the pinacolyl group seems to be a characteristic feature of soman-labeled peptides when they are fragmented in the mass spectrometer, as the soman-labeled butyrylcholinesterase peptide also lost the pinacolyl group during the collision process (32, 33). The 940.7 m/z parent ion in Figure 2 included the pinacolyl group of soman. This is a significant point because it demonstrates that soman had not lost the pinacolyl group while bound to intact albumin protein or during digestion and HPLC separation. The pinacolyl group only dissociated when the peptide was subjected to collision-induced dissociation in the
424 Chem. Res. Toxicol., Vol. 21, No. 2, 2008
Li et al.
Figure 2. Product ion spectrum of the soman-labeled human albumin peptic peptide. The doubly charged parent ion at 940.7 m/z yielded fragments consistent with the sequence VRY*TKKVPQVSTPTL, where * indicates soman covalently bound to tyrosine 411. The accession number for human albumin is GI:28592. In this resource, Tyr 411 is given as Tyr 435 because the numbering begins with the 24-amino acid signal peptide.
mass spectrometer. The presence of the pinacolyl group in the parent ion demonstrates that soman did not age when bound to albumin. At least four separate ion series from the VRYTKKVPQVSTPTL peptide could be extracted from the spectrum in Figure 2. Only the y and b ion series are indicated in the Figure for the sake of clarity. The peptide [VRY*T]KKV[PQ] was seen as part of the b-ion series. The masses of all five observed fragments were consistent with the indicated amino acid sequence plus 78 amu for the added mass of a methyl phosphonic acid. The only residue in this sequence capable of forming a methyl phosphonic acid adduct is the tyrosine, establishing this as the modified amino acid in the peptide. The presence of a 78 amu added mass on the observed fragments, rather than 162 amu, indicates that the pinacolyl group is dissociated from the peptide, in the collision chamber, prior to the onset of backbone fragmentation. The peptide [PQ][VS]TP[TL] was seen as part of the y-ion series. The masses of the observed fragments were consistent with only the amino acids, unencumbered with any adduct mass. Thus, the potentially phosphonylated serine was not labeled, supporting the assignment of the tyrosine as the modified amino acid. The complete amino acid sequence of peptide VRY*TKKV PQVSTPTL was represented by these two ions. These results confirm the conclusion that soman covalently binds to Tyr 411 of human albumin. 3.2. Kinetics for the Phosphonylation of Albumin by Soman and for the Reactivation of Phosphonylated Albumin. 3.2.1. Phosphonylation of Albumin by Soman. Because of the low reactivity of albumin with esters, aryl amides (1, 3),
and DFP (10), a low soman phosphonylation rate was expected. Therefore, the phosphonylation of albumin by soman was carried out under second-order conditions (3, 34), that is, the initial concentrations of soman, [S]0, and albumin, [A]0, were not very different. As a result, the concentrations of the uncomplexed forms of both reactants varied with time during the reaction. Albumin (0.78 mM) was phosphonylated by sub and superstoichiometric amounts of racemic soman (0.32 to 1.3 mM) that inhibited albumin’s aryl acylamidase activity. Reaction was monitored by following albumin’s aryl acylamidase (AAA) activity. Inspection of the inhibition time courses (Figure 3) indicated that less albumin was inhibited by soman than would have been expected from a stoichiometric reaction with the total amount of soman in the reaction. For example, in the presence of a 2-fold excess of soman, 100% inactivation of the AAA activity of albumin was never reached. The titration plot (not shown) for 0.78 mM albumin incubated with different concentrations of soman (0.38, 0.65, 0.81, and 1.30 mM) confirmed this observation. This suggests either that a large part of soman spontaneously hydrolyzed in the nucleophilic Tris/HCl buffer at pH 8.0 or that albumin reacted stereoselectively with specific stereoisomers of soman. 3.2.2. Determination of Fluoride Released during the Reaction of Soman with Albumin. In an effort to decide whether the substoichiometric reaction of soman with albumin was due to a competing spontaneous hydrolysis of soman or to a stereoselective reaction of albumin with soman, ionometric measurements of fluoride released from soman in the presence and absence of albumin were performed (Figure 4). It was found that spontaneous hydrolysis of soman in 60 mM Tris/HCl buffer at pH 8.0 at 25 °C was high. The rate constant for hydrolysis
Reaction of Albumin with Soman
Figure 3. Progressive inhibition of the AAA activity of FAF-HSA ([E] ) 0.78 mM) by different concentrations of racemic soman (b, 0.324 mM; O, 0.520 mM; π, 0.780 mM) in 60 mM Tris/HCl buffer at pH 8.0 at 25 °C. A plot of % residual AAA activity of albumin vs time. The progress curves reach a plateau after consumption of the reactive soman.
Figure 4. Time course of fluoride ions released from soman, in the presence and absence of FAF-HSA. Soman ([S]0 ) bO, 0.52, 24, 0.78, 90, 1.3 mM) was incubated with albumin ([E] ) 0.78 mM) (filled symbols), or soman was incubated by itself through spontaneous hydrolysis (open symbols on continuous curves). Both reactions were in 60 mM Tris/HCl at pH 8.0.
of soman (kh0) was 0.00342 ( 0.00007 min-1. In the presence of albumin, the rate of fluoride release was significantly higher. Semiquantitative analysis of progress curves shows not only that spontaneous hydrolysis of soman occurred at rates comparable to the phosphonylation rates of albumin but also that albumin itself provides a physicochemical environment that enhances the spontaneous hydrolysis of soman. The following is an example of this semiquantitative analysis. The filled squares in Figure 4 describe the release of fluoride after mixing 1.3 mM soman with 0.78 mM FAF-HSA. After 100 min of incubation, 0.3 mM albumin was inhibited (see Figure 3) so that the phosphonylation of albumin accounts for 0.3 mM of released fluoride. Over this time period, about 0.9 mM fluoride was released. Spontaneous hydrolysis of 1.3 mM soman (0 in Figure 4) accounts for about 0.35 mM of released fluoride. This leaves about 0.25 mM fluoride unaccounted for. Thus, the release of this extra fluoride suggests that the hydrolysis of soman at pH
Chem. Res. Toxicol., Vol. 21, No. 2, 2008 425
8.0 is faster in the presence of albumin than in buffer alone. However, dephosphonylation of the Tyr411 adduct is far too slow (0.0044 ( 0.0008 h-1; see section 3.2.5) to account for this albumin-dependent hydrolysis of soman. Thus, this suggests that this catalysis takes place at a different site. However, only one soman adduct on albumin was detected by MALDI-TOF. This finding is consistent with reports on the phosphonylation of bovine serum albumin by FP-biotin (18) and the phosphorylation of human serum albumin by diisopropylfluorophosphate (9). Therefore, it may be hypothesized that albumin-mediated hydrolysis of soman occurs via a mechanism that does not involve a covalent intermediate. Nevertheless, this catalytic effect is extremely weak, about 2 times higher than spontaneous hydrolysis, and may be tentatively regarded as a nonenzymatic, physicochemical interfacial reaction. However, we cannot rule out the existence of a second active center. 3.2.3. Is There Enantioselectivity in the Reaction of Albumin with Soman? Racemic soman is an equimolar mixture of four stereoisomers. Though experiments on the release of fluoride argue against the enantioselectivity of albumin for certain soman isomers, the possibility that albumin’s reaction with soman is stereoselective was tested directly using 31P NMR. Each of the intact isomers of soman (Ps- and Pr-) generates a doublet in 31P NMR (Figure 5, panel A). Soman completely hydrolyzed in 5 N sodium hydroxide gives methylphosphonate (MP) (Figure 5, panel B); the hydrolysis of soman at pH 8.0 gives pinacolyl methylphosphonate PMP (Figure 5, panel D) (35). The reaction between 1.3 mM racemic soman and 0.78 mM albumin was monitored over 24 h by recording 31P NMR spectra. Initial spectra showed the progressive disappearance of the intact soman peaks with the concomitant appearance of three broader resonance peaks, of 80 Hz line width (Figure 5, panel C). After 2 h of incubation, signals for intact soman had vanished. At this point in the reaction, the spectrum of the products was compared to a spectrum of methylpinacolyl phosphonate (MPP). MPP, created under mild conditions of hydrolysis, showed a peak at 23.74 ppm. That peak was unaffected by the presence of albumin (Figure 5, panel D). The most prominent peak in the soman/albumin reaction mixture (the 23.74 ppm peak, f, in Figure 5, panel C) can therefore be assigned to MPP. For longer reaction times (up to 24 h), the MPP peak increased at the expense of the other two peaks in Figure 5, panel C. This strongly suggests that peaks e and d (at 25.59 and 27.56 ppm) correspond to the Ps- and Prcovalent adducts of soman bound to albumin and that there is a slow, spontaneous release of MPP from the of phosphonlyated albumin. Such a release can be described as reactivation. The fact that after 2 h of reaction between albumin and soman, adduct peaks d and e have the same intensity indicates that albumin reacted with the same probability with both the Psand Pr- soman enantiomers. Thus, there is no enantiomeric preference of albumin for the soman isomers. 3.2.4. Effect of MPP on Phosphonylation of Albumin by Soman. Because a large part of soman was hydrolyzed into MPP in Tris buffer at pH 8.0 during the time course of the phosphonylation reaction (Figure 5, panel C), we tested the hypothesis that MPP could compete with soman for noncovalent binding to the Tyr411 site and thereby slow down the phosphonylation reaction. We found that phosphonylation of (0.75 mM) albumin by soman (1.3 mM) in the presence of 1.3 mM MPP caused no change in the progressive inhibition of the AAA activity of albumin. Thus, MPP does not compete with soman. 3.2.5. Spontaneous Reactivation of Soman-Inhibited Albumin Monitored by 31P NMR and MALDI-TOF. Fluoride has been shown to promote dephosphonylation of albumin (36).
426 Chem. Res. Toxicol., Vol. 21, No. 2, 2008
Li et al.
Figure 6. Reactivation kinetics by 31P NMR. Each point represents the integrated area of the NMR peaks (d + e). The areas, in turn, correspond to the relative amounts of adduct. The line is from a fit of the data to eq 1.
Scheme 1
Figure 5. 31P NMR spectra of soman in the presence and absence of albumin. Panel A shows 2 mM racemic soman in dimethylsulfoxided6. The four stereoisomers are indistinguishable. The four resonance peaks correspond to the two doublets of the two pairs of diastereoisomers (35). Panel B shows methyl phosphonate (MP) (a product of soman hydrolysis in 5 N NaOH). Panel C shows 1.3 mM racemic soman after reaction (2 h) with 0.78 mM albumin, in 60 mM Tris/HCl at pH 8; d and e correspond to the PR and PS adducts of albumin, and f is the product of the spontaneous hydrolysis of soman and methyl pinacolyl phosphonate (MPP). Panel D shows 1.3 mM methyl pinacolyl phosphonate (MPP) in the presence of 1.3 mM albumin. The observed peak is that of uncomplexed MPP (35).
However, because of the low nucleophilicity of fluoride ion, it seems unlikely that reactivation was mediated by the fluoride ions released during the phosphonylation step ([F-] e 1.3 mM for the highest soman concentration used). Following the reactivation rate by measuring the recovery of the AAA activity was not accurate; therefore, 31P NMR spectral kinetics was used instead. FAF-HSA (1.3 mM) was inhibited by 1.3 mM soman (10% isopropanol, final). We followed the decrease of the integrated area of the NMR peaks corresponding to the soman-albumin adduct (peaks d + e in Figure 5, panel C) from 10 to 60 h (Figure 6). Because no soman was left after 10 h of incubation, because of its spontaneous hydrolysis (S f S′′ + F-) and to its reaction with albumin (see Scheme 1), the change in the concentration of the soman-albumin conjugate (A-S′) was due only to reactivation (kr). It was assumed that the process was uncompromised and followed pseudo-first-order kinetics (eq 1):
[A - S′]t ) [A - S′]10h · e-kr·(t-10)
(1)
The relative intensity at t ) 10 h was 0.54 ( 0.01; this is equivalent to 0.700 ( 0.015 mM of conjugate. Fitting the data from Figure 6 to eq 1 provided the reactivation constant, kr )
0.0044 ( 0.0008 h-1 ) 0.000073 ( 0.000013 min-1 ) 0.74 ( 0.13 week-1. Thus the half-time for decay of the soman-albumin adduct at pH 8.0 and 25 °C is about 1 week. The stability of the soman-albumin adduct was also measured by mass spectrometry. Soman-labeled albumin was digested with pepsin to release labeled and unlabeled Tyr411 peptides after various periods of incubation. The peptide masses and the area of each peak were determined by MALDI-TOF in reflector mode. Figure 7A shows a scan, after 24 h, where the highest peaks at 1992.1 and 1879.0 have soman bound to Tyr411 in peptides LVRYTKKVPQVSTPTL and VRYTKKVPQVSTPTL. At a later time point after the adduct had been incubated at pH 7.4 and 22 °C for 768 h, the unlabeled peptides at 1717.0 and 1830.1 have a higher intensity than the labeled peptides (Figure 7B). The % label on Tyr411 was calculated from cluster areas and plotted in Figure 7C as a function of time. The half-life for decay of the soman-albumin adduct at pH 7.4 and 22 °C was 20 days. The time course in Figure 7C is first order for nearly two half-lives, supporting the proposal that reactivation is a simple first order process. The difference in rate constants obtained from the mass spectrometry and the NMR experiments is consistent with expectation for the difference in pH. 3.2.6. Progressive Binding of Soman to Tyr411 of Albumin Monitored by 31P NMR Kinetics. The complete reaction path of albumin (A) with soman (S) can be depicted by Scheme 1, which includes the competing spontaneous hydrolysis of soman, second-order phosphonylation of albumin, followed by hydrolytic dephosphonylation (reactivation of the AAA activity of albumin). In Scheme 1, kp is the rate constant of phosphonylation, kr is the rate constant of dephosphonylation (spontaneous reactivation of phosphonylated albumin A-S′), and kh is the rate of the
Reaction of Albumin with Soman
Chem. Res. Toxicol., Vol. 21, No. 2, 2008 427
Figure 8. Time course of the reaction between 1.3 mM soman and 1.3 mM albumin monitored by 31P NMR: b, soman; O, adduct; 2, MPP. The points were measured data. The lines are from numerical fitting of the data to eq 2.
Figure 7. Stability of the soman-albumin adduct calculated from mass spectrometer data. Panels A and B are MALDI-TOF spectra of pepsindigested, soman-labeled human albumin after 24 and 768 h incubation, respectively, at pH 7.4 and 22 °C. Panel C is a plot of % Tyr411 labeled by soman, as a function of time of incubation of the soman-albumin adduct. The soman-albumin adduct decays with a half-life of 20 days.
spontaneous hydrolysis of soman. F- is the fluoride leaving group and S′′ the hydrolysis product, methyl pinacolyl phosphonate (MPP). It follows then that the concentration of all species as a function of time is described by a system of differential equations (eq 2) as sollows:
d[A] ) -kp · [A][S] + kr · [AS′] dt d[S] ) -kp · [A][S] + kh · [S] dt d[AS′] d[A] )dt dt d[S′′] ) -kr · [AS′] + kh · [S] dt d[F ] d[S] )dt dt
(2)
Because the concentration of soman and albumin are of the same order of magnitude and because the rates of spontaneous hydrolysis of soman and phosphonylation of albumin are of the
same order of magnitude, it is not possible to make assumptions allowing a simple integration of the system. However, a numerical solution of the system can be obtained using mathematical software such as Mathematica. Following the residual AAA activity of albumin allows only the measurement of d[A]/dt ) -d[AS′]/dt, while measuring the release of F- gives only d[F–]/dt ) d[S]/dt. However, following the reaction kinetics by 31P NMR gives d[S]/dt ) -d[F-]/dt, d[AS′]/dt ) -d[A]/dt, and d[S′′]/dt in one single experiment (Figure 8). kr was measured separately under the same experimental conditions (kr ) 0.0044 ( 0.0008 h-1; see section 3.2.4 and Figure 6). Fitting the numerical solution of the differential equation system against 31P NMR data gave kp ) 15 ( 3 M-1 min-1 and kh ) 0.0076 ( 0.005 min-1. Thus, the rate constant for spontaneous hydrolysis of soman in the presence of albumin (kh) is about 2 times the rate for the spontaneous hydrolysis of soman in the absence of albumin (kh0 ) 0.00342 ( 0.00007 min-1, from section 3.2.2). This value for kh is in agreement with the estimate for kh made from the measurement of [F-] release, as described in section 3.2.2. 3.3. Molecular Modeling. The crystal structure of HSA with numerous ligands has been described (5). A comparison between these structures shows that Val433 located in the pocket containing Tyr411 might adopt two different conformations depending on the occupant of the pocket. These alternate conformations change the diameter of the pocket and may affect the binding of soman in the pocket. Therefore, we docked soman with both representative crystal structures: 2bxd displaying a narrower pocket in absence of ligand and 2bxh displaying a wider pocket that develops when indoxyl sulfate is bound. The four diastereoisomers of soman were docked in both structures. 3.3.1. Noncovalent Docking by Autodock. Key values from the noncovalent docking results are summarized in Table 1. All four stereoisomers of soman bind better to HSA when the pocket is narrowed by the conformational change of Val433 (2bxd). A narrower pocket provides more stabilizing van der Waals interactions between soman and the residues of the pocket. Therefore, it is expected that the actual conformation of HSA when in complex with soman is similar to that of 2bxd. For 2bxd, the estimated affinity constants for all four stereoisomers
428 Chem. Res. Toxicol., Vol. 21, No. 2, 2008
Li et al.
Table 1. Non Covalent Docking of Soman in the Tyr411 Pocket of HSA Performed by Autodock soman isomer
HSA pdb structure
docked energy (kcal.mol-1)
estimated Kd (µM)
PSCR
2bxd 2bxh 2bxd 2bxh 2bxd 2bxh 2bxd 2bxh
-5.69 -5.56 -5.76 -5.48 -5.69 -5.50 -5.73 -5.42
148 167 115 206 147 176 128 226
PSCS PRCR PRCS
are virtually identical: Kd ) 134 ( 16 µM (Table 1). The pocket seems to be large enough to accommodate all enantiomers of soman (Figure 9). Soman binds with the pinacolyl group stacked against the disulfide bridge formed by Cys392-Cys438. The pinacolyl group makes many hydrophobic contacts with the side chains of Val433, Ile388, Ala449, Phe395, Phe403 and the main chain of Gly434. The binding conformation of soman is largely dictated by its carbon stereocenter. Indeed, PSCS and PRCS display the same binding conformation (rmsd ) 0.07 Å), except that the positions of the fluoride and phosphonate oxygen are inverted as a result of the difference in stereochemistry at the phosphorus. The same applies to the PSCR and PRCR stereoisomers (rmsd ) 0.13 Å). This inversion between the fluoride and phosphonate oxygen does not cost much energy of binding because of the lack of specific interactions between these atoms and residues of the pocket. The phosphorus atom is separated from the hydroxyl of Tyr411 by a distance ranging from 6.80 Å (CR enantiomers) to 7.24 Å (CS enantiomers). This distance is not suitable for a nucleophilic attack of Tyr411 on the phosphorus and explains the slow reactivity of soman toward HSA. It looks like soman may randomly react with Tyr411 on its way in/out of the pocket. 3.3.2. Noncovalent Docking by Gold. In order to check the dependency of the results on the software, to test the robustness of the models obtained by Autodock, and also to generate as many conformers as possible, we decided to repeat the docking simulations using another software, Gold, version 3.1. Gold gave us the possibility of using two scoring functions, Goldscore (force-field-based like Autodock) and Chemscore (empirical, i.e., optimized on a test set of protein–ligand complexes). We performed the docking simulations using both the 2bxh and the 2bxd protein structures. Our results did not show any significant
Figure 9. Models of soman noncovalently docked into albumin. The various stereoisomers of soman were docked in the estero-amidase active center (Tyr411) of HSA using Autodock software.
difference between the structures; therfore, we describe here only the results obtained with the minimized 2bxh protein. We docked the four enantiomers of soman with the minimized 2bxh protein using Gold/Goldscore. The best conformers obtained were similar to the ones obtained by Autodock. The pinacolyl moiety is located at essentially the same place obtained by Autodock (Figure 9); the difference is only a small translation of Tyr411 in the direction of the phosphorus atom (P-OHY411 distances vary from 6.37 to 6.67 Å). As described for Autodock, the rotation of the P (O, F, Me) moiety is relatively free; this is not surprising because of the lack of specific interactions of these atoms with residues of the active site pocket. Again, the phosphorus atom is at a position that does not permit easy phosphonylation of Tyr411. The scores obtained were in the same range for all enantiomers: PSCS ) 39.93, PSCR ) 40.09, PRCR ) 41.28, and PRCS ) 38.16. We obtained very similar results in terms of conformers and energies when we used Autodock or Gold with either 2bxh or 2bxd, which clearly shows the robustness of the models in Figure 9. The Chemscore function is empirical. It can potentially generate different binding modes because contributions of intermolecular interactions are not the same as those with Goldscore or Autodock. The docking using Gold/Chemscore gave different kinds of conformers than Goldscore, but with very small changes in energy. Within the different clusters, some models displayed a phosphorus atom close to Tyr411 (P-OHY411 distances: PRCR ) 4.33 Å, PRCS ) 3.51 Å, PSCS ) 3.98 Å, and PSCR ) 3.41 Å). The enantiomers with PR configuration have a P-F in a favorable position for a nucleophilic attack (OHY411-P-F angle: PRCR ) 155°, PRCS ) 138°, PSCS ) 42°, and PSCR ) 45°). Chemscore docking simulation generated a prephosphonylation complex (fluorine opposite OHY411 and hydrogen bonding between P ) O and OGS489) for the two PR enantiomers. These results suggest that the prephosphonylation step would be energetically more favorable for the PR than for the PS enantiomers. To compare these models with the previous ones, we rescored the prephosphonylation complexes with conformers PRCR and PSCS by the Goldscore function and obtained 10.78 and 27.07, respectively. Compared to the Gold/ Goldscore results (41.28 and 38.16), the rescoring clearly indicates that the ligand positions in the prephosphonylation complexes are not energetically favorable. Therefore, the Chemscore results do not contradict the previous conclusions obtained from the Autodock and Gold/Goldscore dockings. These docking results may explain why the reactivity of soman with HSA is so poor. First, the binding affinity is weak 〈Kd〉 ) 134 µM. Second, the phosphorus does not bind in a position that is favorable for reaction with Tyr411. The docking results also suggest that a higher bimolecular rate constant is probable in the case of PR enantiomers. But because of the poor reactivity and the limited measurement accuracy, an increased reactivity of the PR enantiomers was not experimentally observed. 3.3.3. Covalent Docking by Gold. The noncovalent docking of soman to the HSA protein showed a low enantioselectivity. In order to confirm this finding, we covalently docked the four enantiomers of soman to the 2bxh form of albumin and compared energies (or scores). We used the software Gold, version 3.1, which permits covalent docking. As done before, dockings were scored by the functions Goldscore and Chemscore. To discriminate the conformers obtained by Gold/ Goldscore or Gold/Chemscore dockings, we calculated the energies of soman-protein complexes using mopac-pm3, a
Reaction of Albumin with Soman
Chem. Res. Toxicol., Vol. 21, No. 2, 2008 429
Table 2. Covalent Docking of Soman in HSA Performed by Golda soman isomer
Goldscore/∆Hf
Chemscore/∆Hf
PSCR PSCS PRCR PRCS
11.68/489.8 16.88/483.2 15.67/529.0 12.88/480.4
15.95/519.3 17.43/514.7 16.46/497.8 14.55/539.8
a Scores (Goldscore and Chemscore) have no units. Enthalpies of formation (∆Hf) are in kcal·mol-1.
Figure 10. Models of soman covalently docked into albumin. Various enantiomers of soman were docked into Tyr 411 of HSA (form 2bxh) using Gold software, version 3.1. PSCS (blue), PSCR (red), PRCR (magenta), and PRCS (green).
semiempirical method, and selected those with the lowest ∆Hf. The results are summarized in Table 2. The four models are superimposed in Figure 10. The position of the tertiary butyl group is well conserved for all enantiomers (it points to Leu387, Ile388, Asn391, and Phe403), while the P ) O points to Leu430 and Leu457 for PSCR or PSCS enantiomers, Ser489 (hydrogen bonding with OH) for PRCR, and R410 for PRCS. The conformers PSCS and PSCR are nearly perfect mirror images, the only difference being the position of the CH3 bond to the asymmetric carbon that points to Leu430 for CR and Arg410 for CS. In the case of PR, there are more differences between the two enantiomers because the PRCR enantiomer is the only one that allows hydrogen bonding with Ser489 without having too many disfavored van der Waals interactions. (Arg410 constrains the position of the CH3 bond to C*.) The mopacpm3 ∆Hf values are in the same order of magnitude (∼490 ( 10 kcal.mol-1), showing no thermodynamic enantioselectivity.
4. Discussion 4.1. Roles for Albumin and Albumin/Soman Adducts in the Monitoring of and Protection against Soman Exposure. 4.1.1. Scavenging OP. The reaction of albumin with soman was found to be slow (kp ) 15 ( 3 M-1 min-1 at pH 8.0). This is in agreement with the slow rate of reaction of albumin with the nerve agent simulant, DFP (75 M-1 min-1 at pH 8.2) (10). Combined with the high concentration of albumin present in plasma (0.6 mM), the apparent pseudo-first-order rate constant for the reaction of albumin with soman would be 0.009 min-1 (t1/2 ) 77 min). Thus, despite its high concentration in blood, stoichiometric scavenging of soman by albumin is not expected to play a major role in protection against acute poisoning. Regardless of the low reactivity, significant amounts of OPs are bound to albumin upon in vivo exposure. For example, when
mice were injected with small amounts of the organophosphonate, FP-biotin (10-fluoroethoxyphosphinyl-N-biotinamido pentyldecanamide), amounts that produced no toxic signs, albumin was the most abundantly labeled protein in plasma (46), making it the dominant stoichiometric scavenger. In vitro pretreatment of plasma with a variety of organophosphates (malaoxon, paraoxon, chlorpyrifos oxon, methyl paraoxon, dichlorvos, diisopropylfluorophosphate, diazoxon, and echothiophate) reduced the subsequent reaction of FP-biotin with albumin, indicating that all of these other OPs also react with albumin (46). This strongly suggests that albumin will also be a scavenger for these OPs in vivo. Although albumin displays a low reactivity toward OPs compared to that of cholinesterases and other known secondary biological targets of OPs (37), the large quantity of albumin in blood circulation and lymph may provide a substantial reservoir for sequestering OPs. We also found that phosphonylated albumin slowly self-reactivates. Though the catalytic OPase activity is associated with HSA, and Tyr411 is slow and thus toxicologically insignificant, it may be hypothesized that mutants of albumin capable of hydrolyzing OPs at high rate could be used for the detoxification of soman in vivo. 4.1.2. Biomarker. Phosphonylated albumin may prove to be a useful biological marker of soman exposure (11). The sensitive mass spectrometry methods used by Black et al. (11) and further developed in the present article could prove useful for the diagnosis of soman exposure. In particular, it has been shown that soman tyrosine adducts can be detected at least up to 7 days post exposure in guinea pigs (38). Advantages of a soman-albumin biomarker include its stability and the absence of aging. The half-life for spontaneous reactivation is estimated to be about 1 week at pH 8.0 and 25 °C, and 20 days at pH 7.4 and 22 °C, making the soman-albumin adduct suitable for retrospective studies. The absence of aging means that soman exposure can be unambiguously identified because the somanalbumin adduct does not lose the diagnostic pinacolyl group. Our finding that soman-albumin adducts do not age confirms the reports that there is no loss of alkoxy groups from soman-, sarin-, or DFP-albumin adducts (11, 19, 36). In addition, it has recently been shown that OP-albumin conjugates do not appear to be reactivatable by oximes used as antidotes of OP poisoning. Therefore, the detection of Tyr411 adducts in plasma may be possible, even though individuals exposed to OPs have received effective oxime therapy (38). Another potential method of monitoring for soman exposure is through the use of antibodies. The information presented in this work raises interesting possibilities regarding the generation of antibodies to the soman-albumin adduct. The noteworthy aspect of the soman-albumin adduct in this regard is that Tyr411 is located on the surface of albumin. A soman-albumin adduct at Tyr411 therefore becomes a viable target for antibodies, unlike the soman adducts of cholinesterases, which are buried in a deep gorge well below the enzyme’s surface. In addition, the surface location for the soman-albumin adduct suggests that people exposed to soman might normally produce antibodies against the soman-albumin adduct. This in turn suggests the possibility of monitoring exposure to soman by looking for those antibodies, long after the exposure incident has passed and long after the antigen has disappeared. 4.2. Perspectives: Would Mutant Albumins Be Good OP Scavengers? 4.2.1. Stoichiometric Scavenging. The bimolecular rate constant for the phosphonylation of albumin, kp, 15 M-1 min-1, is about 6 × 106 times slower than that for the inhibition of acetylcholinesterase and butyrylcholinesterase by
430 Chem. Res. Toxicol., Vol. 21, No. 2, 2008
Li et al.
soman (∼9 × 107 M-1 min-1). Because the concentration of soman in the most severe cases of poisoning would be far less than that of albumin in plasma, for example,
